[Bug breakpoints/27422] New: Warn when degrading hw -> sw watchpoint

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27422

            Bug ID: 27422
           Summary: Warn when degrading hw -> sw watchpoint
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: breakpoints
          Assignee: unassigned at sourceware dot org
          Reporter: vries at gcc dot gnu.org
  Target Milestone: ---

Consider test-case test.c:
...
$ cat -n test.c
     1  #include <stdio.h>
     2  #include <stdlib.h>
     3  volatile int a = 1;
     4  volatile int b = 1;
     5       
     6  int main (void)
     7  {
     8    volatile unsigned long long int i;
     9    a = 2;
    10    for (i = 0; i < 1000000000; ++i)
    11      {
    12        b = 2;
    13        b = 3;
    14        b = 4;
    15        b = 5;
    16        b = 6;
    17      }
    18    int *ap = malloc (sizeof(int));
    19    *ap = 1;
    20    printf ("%p\n", ap);
    21    *ap = 0;
    22    a = 0;
    23    return 0;
    24  }
...

Compiled with debug info:
...
$ gcc -g test.c
...

It takes a bit to execute:
...
$ time ./a.out
0x2252260

real    0m2.192s
user    0m2.188s
sys     0m0.001s
...

By disabling ASLR, we see that the malloc'ed pointer doesn't seem to change:
...
$ setarch $(uname -m) -R ./a.out
0x602260
$ setarch $(uname -m) -R ./a.out
0x602260
...

Now consider the following debug session:
...
$ gdb -q a.out
Reading symbols from a.out...
(gdb) b 20
Breakpoint 1 at 0x4005c5: file test.c, line 20.
(gdb) r
Starting program: a.out 

Breakpoint 1, main () at test.c:20
20        printf ("%p\n", ap);
(gdb) watch *(int *)0x602260 == 0
Hardware watchpoint 2: *(int *)0x602260 == 0
(gdb) c
Continuing.
0x602260

Hardware watchpoint 2: *(int *)0x602260 == 0

Old value = 0
New value = 1
main () at test.c:22
22        a = 0;
(gdb) info break
Num     Type           Disp Enb Address            What
1       breakpoint     keep y   0x00000000004005c5 in main at test.c:20
        breakpoint already hit 1 time
2       hw watchpoint  keep y                      *(int *)0x602260 == 0
        breakpoint already hit 1 time
(gdb) 
...

So far, so good.  We set a watch point. It was reported as a hardware
watchpoint when we set it, and it's still a hardware watchpoint when it's
triggered.

Now we restart, and ... this takes a long time, so we press ^C to know what
happened:
...
(gdb) r
The program being debugged has been started already.
Start it from the beginning? (y or n) y
Starting program: /home/vries/gdb_versions/devel/a.out 
^C
Program received signal SIGINT, Interrupt.
_dl_map_object_from_fd (name=name@entry=0x400379 "libc.so.6",
origname=origname@entry=0x0, fd=-1, fbp=fbp@entry=0x7fffffffd3e0,
realname=<optimized out>, loader=loader@entry=0x7ffff7ffe110, l_type=<optimized
out>, mode=<optimized out>, stack_endp=<optimized out>, nsid=<optimized out>)
at dl-load.c:1364
1364      if (__glibc_unlikely (GLRO(dl_profile) != NULL)
(gdb) info break
Num     Type           Disp Enb Address            What
1       breakpoint     keep y   0x00000000004005c5 in main at test.c:20
2       watchpoint     keep y                      *(int *)0x602260 == 0
...

The hw watchpoint silently degraded to a software watchpoint (due to being set
on unmapped memory).

It could be useful to warn in such a case, because a software watchpoint may be
much slower.

-- 
You are receiving this mail because:
You are on the CC list for the bug.